Cinestyle and getting the most out of your DSLR's dynamic range

E

eco_bach

Well-known member
Some have claimed that Technicolor Cinema-style flat-profile ..' will severely limit your dynamic range! Blacks are lifted and whites are crushed'


In pursuit of 'truth' on this issue I invite people to read Mike Seymour's post over at fxphd. He suggests Technicolor's goal was to reduce noise by remapping the sensor's input which makes sense to me http://www.fxphd.com/blog/squeezing-the ... xa-signal/
I'm still confused however on this debate. Anyone care to chime in?
 
ok, some people seem to use "dynamic range" to refer to the size of the color space in which the camera stores the images, and that makes no sense to me - I use dynamic range to refer to the difference in real world brightness between the brightest highlight and the darkest shadow that the camera can record -without clipping- at the same time, and in that sense the way in which that information is stored doesn't matter (whether black is 16 and white is 235, or they are 0 and 255, or 0 and 1, doesn't matter at all for how much dynamic range is captured - for sure we should worry about the number of shades between full black and full white -i.e. the depth of the color space- but that has very little to do wit dynamic range, as I understand it) ----------- having said that, CineStyle does increase the recorded dynamic range a bit over traditional profiles (e.g. Neutral with contrast=-4), but not by much, about two thirds of a stop at most (if you've seen the footage and argue that you see a lot more into the shadows, I'd suggest your screen is crushing the darkest 6% of tones, but the information is there, just look at the waveform) - and in the case of CineStyle this comes at a high price: the inefficient use of the codec's color space (leaving whole chunks of it untouched: 0-15, and usually also 235-255 and most of the color channels) makes the final images more noisy than needed - if you want the same DR as CineStyle but with nicer skin tones and less noise in the final images, I'd suggest my Flaat_2 picture style - if you want as much DR as you can get, you can use my more extreme options, Flaat_3 and Flaat_4, but beware: this increased DR comes from lifting the shadows so they are not clipped, and these no-longer-clipped shadows will display quite a lot of noise in most situations (basically, unless there's a decent amount of light coming from them, which is not easy because, well, they're deep shadows) - more here http://www.similaar.com/foto/flaat-picture-styles/index.html - and here (ckech out my lengthy comment there about my claim of extended DR): http://www.vimeo.com/similaar/flaat
 
NormanBates said:
ok, some people seem to use "dynamic range" to refer to the size of the color space in which the camera stores the images, and that makes no sense to me - I use dynamic range to refer to the difference in real world brightness between the brightest highlight and the darkest shadow that the camera can record -without clipping- at the same time, and in that sense the way in which that information is stored doesn't matter (whether black is 16 and white is 235, or they are 0 and 255, or 0 and 1, doesn't matter at all for how much dynamic range is captured - for sure we should worry about the number of shades between full black and full white -i.e. the depth of the color space- but that has very little to do wit dynamic range, as I understand it) ----------- having said that, CineStyle does increase the recorded dynamic range a bit over traditional profiles (e.g. Neutral with contrast=-4), but not by much, about two thirds of a stop at most (if you've seen the footage and argue that you see a lot more into the shadows, I'd suggest your screen is crushing the darkest 6% of tones, but the information is there, just look at the waveform) - and in the case of CineStyle this comes at a high price: the inefficient use of the codec's color space (leaving whole chunks of it untouched: 0-15, and usually also 235-255 and most of the color channels) makes the final images more noisy than needed - if you want the same DR as CineStyle but with nicer skin tones and less noise in the final images, I'd suggest my Flaat_2 picture style - if you want as much DR as you can get, you can use my more extreme options, Flaat_3 and Flaat_4, but beware: this increased DR comes from lifting the shadows so they are not clipped, and these no-longer-clipped shadows will display quite a lot of noise in most situations (basically, unless there's a decent amount of light coming from them, which is not easy because, well, they're deep shadows) - more here http://www.similaar.com/foto/flaat-picture-styles/index.html - and here (ckech out my lengthy comment there about my claim of extended DR): http://www.vimeo.com/similaar/flaat
Click to expand...

Quantization does figure into 'representation dynamic range'. In the extreme example of two bits of quantization, usually mapped to 'black on the screen' and 'white on the screen', your 'dynamic range' is 2. Whatever the real world scene may have had, is mapped into those two values...

In the case of representation into 256 values, one has 8 stops (that is 8 doublings) of 'range' of the scene, if there is perfect quantization of the scene intensities. As it is, the noise of the sensor figures into things, so one does not get 0-255 'values', and so the representation has a less than 8 stop dynamic range.

While the actual human viewable scene may be many be in the thousands to 1 range... the camera + digitizer will only represent that range in terms of the bit depth of quantization.

Some of the popular ways to get around this are 1) non-linear digitization, 2) deeper bit depth. More 'advanced' techniques would involve being able to control the photo 'cell' and it's repsonse to light selectively.

For still photography, this would be like automatically taking 2 or 3 shots of the same scene, all at say 8 bits of quantization, but adjust the exposure for 'low light/shadow', 'mid tone', 'highlight detail'. Then meld the resulting images into an image that has say 24 bits of representation. That way whatever the original scene may have been in terms of 'dynamic range', the resulting digital image better represents that range in terms of discernible grey values (or 3x24 for color...).

For moving picture digitization, reading out the sensor 'data' at different times, without 'resetting the cell', would allow for the high intensity values to be read out before they saturate the cell, where as low intensity areas would be allowed to 'accumulate'. This would have an impact on the effective shutter speed of the camera... but heck... its an idea...

In the case of non-linear representation, that has been done for years with audio, and essentially 'compresses' the input dynamic range, then on output, expands the compressed signal.
 
I shoot almost exclusively in Technicolor cinestyle, mainly for the lifted shadows. What it effectively does is allow me to choose the crush point. Sometimes I may crush them right back down to where they would have been in a more conventional profile, sometimes I might retain some of that detail cinestyle exposed. I can choose that shot by shot. Skin tones can be issues, but not bad if the face is properly exposed. I expose/frame/focus in a faux Kodachrome profile that is super contrasty and super saturated with a red bias, then switch to cinestyle right before I roll.
 
Last edited:
I think this anti-Cinestyle sentiment comes from those who don't really understand the issues (and I think that the appreciation of one style over the other is largely a matter of personal preference). How can the engineers at Technicolor, working in conjunction with Canon, have fouled up like that? It doesn't pass the smell test, first of all. And how many of these competing styles are being developed by people who have equal experience with Techicolor engineers?
 
Gary Huff said:
I think this anti-Cinestyle sentiment comes from those who don't really understand the issues (and I think that the appreciation of one style over the other is largely a matter of personal preference). How can the engineers at Technicolor, working in conjunction with Canon, have fouled up like that? It doesn't pass the smell test, first of all. And how many of these competing styles are being developed by people who have equal experience with Techicolor engineers?
Click to expand...

While I'm not a Canon user, I think my thoughts are general enough to hold... What I want in terms of capture is the mode that allows for the data to be recorded with the least processing, and the most 'signal'. In still cameras and high end digital film cameras, the concept of 'raw' is approaching that ideal.

The other part of the problem is that one can not, at least looking at the easy to find Technicolor pages for Cinema Style for Canon... one can not find any technical documentation on exactly what the 'style' does... other than a few code words such as s-log representation... and 'footage will look 'flatter'', and will require specialized LUT's or processing to be used in the post processing to finish with acceptable contrast...

Now that sounds fine to me... but for someone who is use to seeing Kodachrome saturated color, contrasty images, out of their box... they may think 'something is wrong'...
 
I obviously have less experience than Technicolor engineers
and still I think Flaat_2 delivers cleaner images
it's up to you to test it and decide for yourself - it's free

--

@j1clark@ucsd.edu
I totally get what you mean
e.g. about "0 and 1", I know you can just represent 2 stops with 2 values, I was thinking of say, 32 bits of floating point niceness packed between 0 and 1
in any case, to me, as long as I have, say, 128 values, I wouldn't say this affects dynamic range, I'd say it affects gradation
anyway...

--

now, on the DR issue, let me post here some waveforms from my DR test chart (which works on thirds of a stop)

this is Portrait with contrast=-1
portrait_1.jpg


Neutral with Contrast=-4 brings up about a third of a stop higher DR:
neutral_4.jpg


Marvels Cine brings up the shadows a bit more:
marvels_3.jpg


and CineStyle even a little bit more (also, it rescales everything, so it only uses 16-256, and in practice just 16-212 except for the brightest two-thirds of a stop, but this doesn't affect how dark the darkest non-clipped shadows are)
technicolor_4.jpg


Seeing this, and noticing that nobody has had any success trying to revive any detail in the highlights, I deceded to spread the image more evenly across the codec's color space, hence Flaat_2
flaat_2.jpg


BTW this was also inspired by what Sony S-Log does:
abelcine_slog.jpg


and now, if you want more DR, Flaat_3 and Flaat_4 can give it to you, but the heavily-pushed-up shadows, while no longer clipped, will usually be noisy:
flaat_4.jpg
 
Last edited:
I echo the sentiments of Gary. Technicolor worked WITH Canon, and was given access to the image data higher up the signal chain than any user has the ability to.
Samuel, I applaud your effort in creating your flaat styles, and although i haven't tried them, can't see from your posted examples how they improve upon Cinestyle. BTW, also like your 3rd party lens reco's though would add the Tammy 28-75 to the mix.
 
dwyz said:
Omg charts!
Click to expand...

Yeah, for me 'charts' of step wedges answer a whole bunch of 'practical' questions, in a few comparison glances... Unfortunately I don't have the money for 'real' step wedges, but fortunately, Barry Green posted some for the GH-1 a long time ago, which clarified a number of things for the various modes that camera could shoot at, and which set of settings I've been using since.

Next up would be a Gretag-Macbeth color chart, or similar, and how much 'color' futzing is done with the various mode selections...
 
j1clark@ucsd.edu said:
Yeah, for me 'charts' of step wedges answer a whole bunch of 'practical' questions, in a few comparison glances... Unfortunately I don't have the money for 'real' step wedges, but fortunately, Barry Green posted some for the GH-1 a long time ago, which clarified a number of things for the various modes that camera could shoot at, and which set of settings I've been using since.

Next up would be a Gretag-Macbeth color chart, or similar, and how much 'color' futzing is done with the various mode selections...
Click to expand...

Barry's articles are great. Figured there must be accurate but affordable calibration-step wedge chart available.
A quick google search turned up these...most seem WAY overpriced but will probably pick up a basic Kodak step wedge as a start to do my own tests.

http://www.imatest.com/docs/q13/#aicharts

http://www.adorama.com/KKQ13.html?u...te&utm_campaign=Other&utm_source=RFLAID016047

http://www.dsclabs.com/colorbar_grayscales.htm

http://www.danes-picta.com/

BHphotovideo also carries the inexpensive Q-13
http://www.bhphotovideo.com/c/product/714596-REG/Tiffen_EK1527654T_Q_13_Color_Separation_Guide.html
 
Last edited:
eco_bach said:
Barry's articles are great. Figured there must be accurate but affordable calibration-step wedge chart available.
A quick google search turned up these...most seem WAY overpriced but will probably pick up a basic Kodak step wedge as a start to do my own tests.

http://www.imatest.com/docs/q13/#aicharts

http://www.adorama.com/KKQ13.html?u...te&utm_campaign=Other&utm_source=RFLAID016047

http://www.dsclabs.com/colorbar_grayscales.htm

http://www.danes-picta.com/

BHphotovideo also carries the inexpensive Q-13
http://www.bhphotovideo.com/c/product/714596-REG/Tiffen_EK1527654T_Q_13_Color_Separation_Guide.html
Click to expand...

Yeah, that's the problem with the official charts... way over priced for the 'occasional' use... I've had a Kodak 18% Grey card for years, I have the Kodak "Professional Photoguide" book as my 'reference' for color... I think I have a 'cheap' GretagMacBeth-esque color chart I got with a expensive (for the era...) scanner quite a few years ago, but I've lost track of it...

I think I like the idea of using an HDMI display and 'making' charts, as NormanBates indicated... not NBS traceable quality... but perhaps good enough to get some thoughts going about what is 'happening' with the various camera modes.

My usual 'focus/lens' quality chart was a newspaper put up on a wall...
 
like the idea of a newspaper, if you can get it flat enough! I think $30 for the KodakQ-13 is a tiny investment for a half decent reflected light calibration reference. Especially useful when comparing picture styles in camera. I would only add though the importance of consistent exposure, and use of a decent incident light meter! Otherwise you're comparing apples and oranges;)
 
You must log in or register to reply here.
Back
Top